ytongbai/LVM

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1838 stars · 61 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

LVM is a large vision model trained with sequential next-token prediction over 'visual sentences', using no linguistic data. It builds on OpenLLaMA for autoregressive modeling and OpenMuse VQGAN for image tokenization, with training code and model definitions from 100M to 30B parameters.

Use cases

• train a large vision model without language data
• pretrain autoregressive models on visual tokens
• run next-token prediction over images and videos
• solve vision tasks with visual prompts
• scale vision model training across model sizes
• convert images into visual tokens with VQGAN

When to choose

• you want to pretrain or fine-tune large vision-only autoregressive models
• you need JAX-based distributed training on GPU or TPU clusters
• you are researching language-free visual representation learning

When to avoid

• you need a ready-made inference API or hosted model for production apps
• you want off-the-shelf image classification or detection rather than research training
• you lack multi-GPU/TPU resources for large-scale training

Facets

library · maturity active

machine-learning deep-learning llm-training machine-learning deep-learning computer-vision artificial-intelligence python vision-model autoregressive visual-sentences vqgan pretraining jax research gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ytongbai/LVM")